Social network analysis on overlapping multiple mailing-lists in a company, H Takeda, S Yamaguchi, S Hara, D Chiba, I Ohmukai

Tags: National Institute of Informatics, Hideaki Takeda, University of Tokyo, team members, integrated network, company Information, team, social network analysis, personal relation, Satoshi Yamaguchi, social network, mailing-lists, Ryutaro Ichise National Institute of Informatics Hideaki Takeda, Hideaki Takeda National Institute of Informatics, Sunbelt XXVII International Sunbelt Social Network Conference, Systems Inc Ikki Ohmukai, organizations
Content: Sunbelt XXVII International Sunbelt Social Network Conference, Chandris Hotel Complex at Dassia Bay, Corfu Island, Greece, May 1-6, 2007 Social network analysis on overlapping multiple mailing-lists in a company Hideaki Takeda National Institute of Informatics, The University of Tokyo Satoshi Yamaguchi, Seiichiro Hara, Daisaku Chiba Alpha Systems Inc Ikki Ohmukai, Ryutaro Ichise National Institute of Informatics Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} Social network analysis on overlapping multiple mailinglists in a company z Abstract: Organization of recent companies becomes flexible so that shortterm project teams are quickly organized and dissolved within them. The social network of employees tends to be multilayered because employees are often members of multiple teams. In such organizations, mailing-lists are useful communication tools within teams because they can be quickly generated wherever members are working. We analyzed multiple mailing-lists in a company to explicate how communication within and cross teams occurred and related to each other. We obtained some observations as a result of the analysis. Central members in the social network in a mailing-list often become central in the social network joined with different mailing lists. Betweenness is more stable than closeness between two social networks. Variation of betweenness reflects closeness in business among projects. Members with high betweenness in the joined social network are often managers in the company. It seems to be the evidence that the management works well. Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} 1
Background z Needs for improvement in knowledge sharing in organizations z Needs for explication of "information ecosystem" in organizations For better efficiency of information distribution For identifying bottlenecks in information sharing Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} Methods z Various communication methods are used now Mails, phones, messengers, face-to-face communications and so on z We select mailing lists (ML) Logs are available Mailing lists in organizations are mostly used for business cf. mails are used for various purposes from private to business Privacy issues are minimum Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} 2
The Nature of mailing lists in the company z The General Information of the target company Information Technology business system development for client companies Team-based The company forms a new team for a new system development. When it is finished, the team is disbanded z The role of mailing lists The primary communication method among the team members Team members are often distributed between the company office and client office Each team maintains at least one mailing list Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} Goal for the analysis z Extract social relationship among members in and cross the teams z Intra-team relationship Who is the key person in team? What is the role of members? z Inter-team relationship Who is the key person in the specific tasks (not teams)? top management often fails to identify employee's experience What is relationship among business tasks? How does employee change their role in different teams? Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} 3
Extraction of personal relation from mailing list logs z A relation from Person B to Person A is identified when Person B posts a follow-up mail to the mail posted by person A z We treat the above directed relations as indirect relations in analysis
Subject: Re: ABC problem > I found a problem ... It is easy to solve ... Person B
Subject: ABC problem I found a problem ... Follow-up post A relation is identified Person A
.........
Number of interaction is considered as weight
Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo}
Details of the target mailing lists z Facts Number of mailing lists: ca. 150 Number of nodes: ca. 1,500 Period: May 2005 to August 2005 (ca. four months) Number of mails: ca. 85,000 z Security treatment Unique number for each mail account Hide personal information Use header part only Avoid secret information like NDA (Non-Disclosure Agreement) Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} 4
A single mailing-list network
Active Users
participant
Non Active Users
Follow-up Relation
By Pajek
Analyze Active Users
Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo}
The whole network Aggregation of multiple mailing lists on the similar business Employee bridging different business areas Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} 5
Bridging ML aggregates z A ML aggregate is a virtual organization z Participants belonging multiple organizations are expected to bridge these organizations z Analysis Analysis on the whole network comparison between network index and office organization Analysis on the single-ML network and the multi-ML network On heavily overlapping MLs On weakly overlapping MLs Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo}
Betweenness centrality and office organization in the whole network
Rank Node ID Betweenness Job Grade
1
92 0.146193345 director
2
642 0.142107724 section head
3
97 0.128738904 director
4
96 0.117601719 director
5
638 0.110096847
6
35 0.091883537 employee
7
20 0.083169606 employee
8
647 0.078591173
9
1154 0.062681324 employee
10
906 0.055225532
11
108 0.054742391 director
12
177 0.051995841
13
14 0.051918424 manager
14
136 0.046871807 section head
15
672 0.041717180 section head
16
418 0.038996085 section head
17
419 0.037065126
18
137 0.034937544
19
427 0.034510085 section head
20
856 0.033964111
21
668 0.033009429
22
115 0.029313637 branch manager
23
1 0.028682580 section head
24
112 0.027642368 section head
25
122 0.026788383 manager
z Top ranks of betweenness centrality z Many management-level employees May be sound organization
z We decide use betweenness centrality for the further analysis
Job Grade
A typical team
director
1
manager
2
section head
10
employee
100
Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo}
6
The role of social networks in ML z In network from a single ML: We can understand Relationship among people in a single business task z In network from multiple MLs: We can understand Relationship among people in multiple tasks Relationship among tasks Changing roles of participants according to tasks Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} Analysis on multi-ML network z Method Compare the single-ML networks with the integrated network in betweenness centrality z Point of analysis Can we see relationship between tasks? Can we find key persons bridging tasks? Can we see how role of a single person change? Can we presume features in original networks from the integrated network? z Three overlapping mailing lists ML1: 228 participants ML2: 295 participants ML3: 287 participants Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} 7
General relation of overlapping mailing lists
No Overlap
Weak Overlap ML1
ML1-ML2
ML1-ML2-ML3
ML1-ML3
ML2
Strong OverlapML2-ML3
ML3
Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo}
Network from weakly overlapping mailing lists
1030 936 1101 ML1 1030
936 1030 1101
1101 936
ML1-ML3
ML3
ML3 Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo}
8
Network from weakly overlapping mailing lists
Mem ID ber ML 944 1 936 1-3 897 1 955 1 881 1 889 1 1063 3-1 1834 3 1030 3-1 1101 3-1 956 3-1
Single ML network
ML1
ML3
Rank
# of Posts
Rank
# of Posts
12
49 -
-
7
73
67
1
1
136 -
-
33
74 -
-
5 1004 -
-
28
38 -
-
23
1 17
75
-
-
10
92
129
8 109
91
118
2 26
53
Merged network ML1-ML3 Rank 8 2 4 61 7 42 5 3 9 6
z May indicate key persons to bridge two MLs z Participants who posted to both MLs tend to rise their rank in betweenness centrality z Participants who posted a single ML decrease their ranks z Even low ranking participants in both networks may rise their rank in the integrated network. If the posted numbers are unbalanced, change is high It indicates difficulty to presume the original networks Section head Section head
153
3 120
7
54
Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo}
Network from strongly overlapping mailing lists
1524 1544 1570 ML2 1524 1544
1570 1524 1544 ML2-ML3
ML3
Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo}
9
Network from strongly overlapping mailing lists
ID
Member ML
1524 2-3 1806 3 1528 2-3 1542 2-3 1544 2-3 1570 2-3 1543 2 1854 3 1531 2-3 1534 2-3 1824 3 1575 2-3
Single ML network
ML2
ML3
Rank
# of Posts
Rank
# of Posts
1
357
5
96
-
-
1 7833
2
286
4
56
6
178 19
39
Merged network ML2-ML3
z Participants who posted to both MLs tend to rise their rank in betweenness centrality z Participants who posted a single
Rank
ML decrease their ranks
z But, the both effects are smaller 1 than "weakly case"
2 It indicates that presumption
3
to the original networks is
easier
7
15
99 36
46
8
3
68 -
-
9
4
220 39
22
10
-
-
8
76
11
20
74 59
13
12
44
122 14
35
17
-
-
11
103
18
40
136 38
47
20
Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo}
Conclusion and future task z Conclusion Mailing list can be used as sources of social network analysis Easy to obtain (often available as public) Analysis of social network from multiple mailing lists Different structures depending on relationship between tasks Exhibit key persons difficult to identify with single networks z Future task From analytical point-of-view Must include other media for communication (chat, mail etc.) From organizational management Positive support for key persons z Stimulation to increase communication Hideaki Takeda @ {National Institute of Informatics, The University of Tokyo} 10

H Takeda, S Yamaguchi, S Hara, D Chiba, I Ohmukai

File: social-network-analysis-on-overlapping-multiple-mailing-lists.pdf
Title: Microsoft PowerPoint - sunbelt07-2.ppt
Author: H Takeda, S Yamaguchi, S Hara, D Chiba, I Ohmukai
Author: takeda
Published: Fri May 11 06:30:51 2007
Pages: 10
File size: 1.45 Mb

Copyright © 2018 doc.uments.com